Worker Training, Firm Productivity, and Trade Liberalization: Evidence from Chinese Firms

Abstract

The present paper discuss a novel mechanism–worker training–that output trade liberalization affects firm productivity. With disaggregated Chinese firm-level production data from 2004 to 2006, we find strong evidence that output trade liberalization boost firm productivity. More importantly, after controlling for firm's self-selection to invest on worker training, our extensive empirical search suggests the following findings.

This chapter is coauthored with.Qing Liu and Larry Qiu and originally published in Developing Economies, 2017, 55(3), pp. 189–210.

Appendix A

Measuring Ex-ante TFP (TFP2)

This section draws heavily from Qiu and Yu (2013) to discuss how we construct and measure TFP using two different approaches: ex-post TFP (TFP1) and ex-ante TFP (TFP2) inspired by Feenstra et al. (2014).

We extend the Olley-Pakes (1996) approach to fit with China's reality in the following ways. Firstly, given that the measure of TFP requires real terms of firm's inputs (labor and capital) and output, we adopt different price deflators for inputs and outputs from Brandt et al. (2012) in which the output deflators are constructed using “reference price” information from China's Statistical Yearbooks whereas input deflators are constructed based on output deflators and China's national input-output table (2002).

Secondly, we take China's WTO accession in 2001 into account since such a positive demand shock would push Chinese firms to expand their economic scales, which in turn can exaggerate the simultaneous bias of their measured TFP. Thirdly, it is essential to construct the real investment variable when using the Olley-Pakes (1996) approach. As usual, we adopt the perpetual inventory method to investigate the law of motion for real capital and real investment. Different from assigning an arbitrary number for the depreciation ratio, we use the exact firm's real depreciation provided by the Chinese firm-level data set.

Finally, we also consider firm's processing behavior in the TFP realization by constructing a processing export indicator (one denotes processing export and zero otherwise). The idea is that processing firms may use different technology than non-processing firms (Feenstra and Hanson, 2005).

Thus, a firm's investment function is

$$ V_{it} = g_{1} (x_{it} ,\ln K_{it} ,EX_{it} ,PE_{it} ,WTO_{t} ) $$

where EX_it,(PE_it) is the export (processing export) indicator to measure whether firm i exports (engages in processing exports) in year t, and WTO_t is an indicator that equals one if the WTO agreement has occurred after 2001 and zero before that. Therefore, inverting the investment function with respect to its first argument we obtain⁹:

$$ x_{it} = g_{1}^{ - 1} (V_{it} ,\ln K_{it} ,EX_{it} ,PE_{it} ,WTO_{t} ) $$

Given the gross production function

$$ \ln Y_{it} = \alpha_{k} \ln K_{it} + \alpha_{l} \ln L_{it} + \alpha_{m} \ln M_{it} + x_{it} + \varepsilon_{it} $$

and defining the function $g_{2} ( \cdot )$ as $\alpha_{k} \ln K_{jt} + g_{1}^{ - 1} (V_{it} ,\,\ln K_{it} ,\,EX_{it} ,\,PE_{it} ,\,WTO_{t} )$, the estimation of the labor (materials) coefficients $\alpha_{m}$ ($\alpha_{m}$) are obtained as:

$$ \ln Y_{it} = \alpha_{l} \ln L_{it} + \alpha_{m} \ln M_{it} + g_{2} (V_{it} ,\ln K_{it} ,EX_{it} ,PE_{it} ,WTO_{t} ) + \varepsilon_{it} $$

The next step is to obtain an unbiased estimated coefficient of $\alpha_{k}$ Olley-Pakes (1996) use the following specification:

$$ \ln Y_{ir} - \hat{\alpha }_{l} \ln L_{it} - \hat{\alpha }_{m} \ln M_{it} = \alpha_{k} \ln K_{it} + E(x_{it} |x_{it - 1} ,pr_{it} ) + [x_{it} - E(x_{it} |x_{it - 1} ,pr_{it} )] + \varepsilon_{it} $$

where the estimated values of the labor coefficient and materials coefficient are used on the left. The expectation of productivity appearing in (2-step) is modeled as a forth-order polynomial function of lagged productivity, which can be obtained as (g_2i,t-1-$\alpha_{k}$ln K_i,t-1), and also the predicted probability of the firm's survival into the year t, pr_it , based on year t-1 information. The predicted probability is obtained from Probit estimation.¹⁰ The term [x_it-E(x_it| x_it-1, pr_it)] is the productivity shock for surviving firms, but does not affect the investment or exit choice so it is treated as an error.

Once the coefficient of capital $\hat{\alpha }_{k}$ is estimated in Eq. (10.2-step), it is ready to obtain the standard ex-post TFP:

$$ TFP1_{it} \equiv x_{it} = \ln Y_{it} - \hat{\alpha }_{k} \ln K_{it} - \hat{\alpha }_{l} \ln L_{it} - \hat{\alpha }_{m} \ln M_{it} $$

In this way, TFP1 includes both true production productivity and managerial efficiency. By contrast, the ex-ante productivity (TFP2) which only capture true production productivity is given by

$$ TFP2_{it} = g_{1}^{ - 1} (V_{it} ,\ln K_{it} ,EX_{it} ,PE_{it} ,WTO_{t} ) $$

Appendix Table 10.A1 reports the estimated coefficients of labor, capital, materials, TFP1 and TFP2.

Table 10.A1 Total factor productivity of Chinese firms (2000–2007)